Mohammad Esmaeilzadeh

Optimizing completion delay in network coded systems over TDD erasure channels with memory

In this paper, we consider a network comprising a sender and a receiver in a time division duplex (TDD) erasure channel with memory, where transmission of network coded packets by the sender and listening to feedback from the receiver (about how many more coded packets are required) occur in rounds. The main objective is to decide the number of transmissions in the next round based on the received feedback such that the expected transmission time of M packets is minimized. Compared to previous studies where only memoryless erasures were considered, we incorporate memory into the optimization framework through a Gilbert-Elliot channel (GEC) model with Good and Bad states at the bit level and packet level. We then utilize infrequent feedbacks about the last channel state to make a more informed decision about the optimum number of coded packets for the next transmission round. Our numerical results show that when the memory content of the channel is relatively high, our scheme can make significant improvement in the mean completion delay (as high as 74%) compared to the scheme which is oblivious to the erasure temporal correlations. We also study the effect of packet length on the delay/throughput optimization in our framework and discuss its implications for the design of practical transmission networks.

Joint Optimization of Throughput and Packet Drop Rate for Delay Sensitive Applications in TDD Satellite Network Coded Systems

In this paper, we consider the issue of throughput and packet drop rate (PDR) optimization as two performance metrics for delay sensitive applications in network coded time division duplex (TDD) satellite systems with large round trip times (RTTs). We adopt random linear network coding (RLNC) and our purpose is to obtain the optimum RLNC-based transmission strategy. We start with a single-user case and propose a systematic framework to investigate the advantage of using feedback by comparing feedback-free and feedback schemes. Showing analytically that the feedback-free scheme gives better performance for our system of interest, we extend it to multi-user broadcast case. To this end, we consider a number of different broadcast scenarios and optimize the system parameters such that the best overall performance is achieved. Furthermore, the complicated interplay of the mean throughputs and PDRs of different users with different packet erasure conditions is discussed. Finally, it is shown that the optimized feedback-free RLNC broadcast scheme works close enough to an idealistic RLNC scheme, where the complete and immediate knowledge about the reception status of all users is assumed to be available at the sender.

Random Linear Network Coding for Wireless Layered Video Broadcast: General Design Methods for Adaptive Feedback-Free Transmission

This paper studies the problem of broadcasting layered video streams over heterogeneous single-hop wireless networks using feedback-free random linear network coding (RLNC). We combine RLNC with unequal error protection (UEP) and our main purpose is twofold: to systematically investigate the benefits of UEP+ RLNC layered approach in servicing users with different reception capabilities and to study the effect of not using feedback, by comparing feedback-free schemes with idealistic full-feedback schemes. To these ends, we study “expected percentage of decoded frames” as a key content-independent performance metric and propose a general framework for calculation of this metric, which can highlight the effect of key system, video, and channel parameters. We study the effect of number of layers and propose a scheme that selects the optimum number of layers adaptively to achieve the highest performance. Assessing the proposed schemes with real H.264 test streams, the trade-offs among the users’ performances are discussed and the gain of adaptive selection of number of layers to improve the trade-offs is shown. Furthermore, it is observed that the performance gap between the proposed feedback-free scheme and the idealistic scheme is very small and the adaptive selection of number of video layers further closes the gap.

Deep phenotyping: deep learning for temporal phenotype/genotype classification

BackgroundHigh resolution and high throughput genotype to phenotype studies in plants are underway to accelerate breeding of climate ready crops. In the recent years, deep learning techniques and in particular Convolutional Neural Networks (CNNs), Recurrent Neural Networks and Long-Short Term Memories (LSTMs), have shown great success in visual data recognition, classification, and sequence learning tasks. More recently, CNNs have been used for plant classification and phenotyping, using individual static images of the plants. On the other hand, dynamic behavior of the plants as well as their growth has been an important phenotype for plant biologists, and this motivated us to study the potential of LSTMs in encoding these temporal information for the accession classification task, which is useful in automation of plant production and care.MethodsIn this paper, we propose a CNN-LSTM framework for plant classification of various genotypes. Here, we exploit the power of deep CNNs for automatic joint feature and classifier learning, compared to using hand-crafted features. In addition, we leverage the potential of LSTMs to study the growth of the plants and their dynamic behaviors as important discriminative phenotypes for accession classification. Moreover, we collected a dataset of time-series image sequences of four accessions of Arabidopsis, captured in similar imaging conditions, which could be used as a standard benchmark by researchers in the field. We made this dataset publicly available.ConclusionThe results provide evidence of the benefits of our accession classification approach over using traditional hand-crafted image analysis features and other accession classification frameworks. We also demonstrate that utilizing temporal information using LSTMs can further improve the performance of the system. The proposed framework can be used in other applications such as in plant classification given the environment conditions or in distinguishing diseased plants from healthy ones.

Inter-session network coding for transmitting multiple layered streams over single-hop wireless networks

This paper studies the problem of transmitting multiple independent layered video streams over single-hop wireless networks using network coding (NC). We combine feedback-free random linear NC (RLNC) with unequal error protection (UEP) and our goal is to investigate the benefits of coding across streams, i.e. inter session NC. To this end, we present a transmission scheme that in addition to mixing packets of different layers of each stream (intra-session NC), mixes packets of different streams as well. Then, we propose the analytical formulation of the layer decoding probabilities for each user and utilize it to define a theoretical performance metric. Assessing this performance metric under various scenarios, it is observed that inter-session NC improves the trade-off among the performances of users. Furthermore, the analytical results show that the throughput gain of inter-session NC over intra-session NC increases with the number of independent streams and also by increasing packet error rate, but degrades as network becomes more heterogeneous.

SPHARM-based shape analysis of hippocampus for lateralization in mesial temporal lobe epilepsy

Spherical harmonics (SPHARM) is a powerful tool for modelling and processing of 3D connected objects of any shape. SPHARM popularity lies in its capability of revealing surfaces global discrepancies in a multi-scale manner. In medical image analysis, this capability is of great importance in diagnosis of diseases that are related to deformations in the brain structures, such as mesial temporal lobe epilepsy that is associated with the hippocampus deformation. In this paper, we present a simple and practical method for SPHARM registration, which is required for conducting shape comparisons. The method utilizes concepts of principal components and solves the challenging problem of SPHARM registration. Our method benefits from characteristics of SPHARM coefficients that have independent [x,y,z] elements; so registration is easily performed in the SPHARM feature space. Then, we propose our feature selection methods that summarize 1536 SPHARM-based features of each subject into three lateralization indices. These three indices measure the distances between left and right hippocampi of healthy and epileptic subjects to detect the epileptogenic hippocampus. This work improves the lateralization accuracy from 78% of conventional volumetric method to 85%, and also in cases where volumetric analysis is uncertain, 16% improvement is achieved. This method could be used as a compliment to other methods to decrease lateralization error.

On Reducing Intercept Probability for Unsubscribed Video Layers Using Network Coding

In this letter, we use random linear network coding (RLNC) to deliver a two layered video sequence to two wireless subscribers such that a high-paying subscriber receives the highest quality and a low-paying subscriber receives the basic quality. We first derive closed-form expressions for the probability of the low-paying subscriber intercepting sufficient coded packets to experience the highest video quality. We consider both feedback-aided and feedback-free transmissions. We propose a resource allocation framework that minimizes the intercept probability subject to deadline and subscribed video quality constraints. We show the tightness between the theoretical and simulation results, and illustrate the protection achieved by RLNC for subscription-based layered video delivery.